DTV Certified, What Should It Mean?
The official start date for terrestrial DTV broadcasting was 1 November 1998. While some of the 1080i horses were out of the gate a few days early with the launch of the space shuttle, in August of 2000 we are still building the track so close to the horses that we seem to be loosing sight of our direction. Way to many organizations involved in DTV are trying to make it simple. It isn’t and won’t be for a good number of years to come. Fair warning, this is history repeating itself. Oversimplifying NTSC lead us right into “Never Twice the Same Color” and it is happening again with DTV.
Check out what’s being sold at retail as DTV compatible or DTV ready. Knowing a bit about the real DTV system, you might find yourself wondering how some of these sets could possibly be connected with digital transmission of component video when they are nothing more than ordinary TV’s. Does an S video input on a 1.33 aspect ratio TV set somehow qualify it for a DTV nod? One could argue that retailers aren’t far off in doing so, after all first generation of set top boxes, such as the Panasonic TU-DST 50 W, provided a down converted S and composite video output. But then the S output was removed from the second generation TU-DST 51 W box. (We are not making a judgment about that move, just stating a fact.)
If current TV sets are good enough to be considered DTV compatible, what added value is there to DTV that could justify the broadcast and program production community’s significant additional investment? Should we not try to create the kind of difference in digital versus analog that existed in the transition between black and white and color? Are broadcasters justified in hinting that CEA numbers for DTV compatible sets sold are seriously inflated? Are the real numbers so low that broadcasters can justify a go slow approach to DTV, or HDTV in particular?
If DTV is to have added value should we not encourage display device manufacturers to recognize that it is imperative to build a dramatically better viewing device?
The ATSC and CEA currently have a DTV Certification program. As we understand it, the major application is towards reception and decoding of the digital signal, not its display. Efforts are being made to include the display in this certification process. Towards that effort we would like to offer our own ideas.
We approach this from the point of view that display devices should first clearly demonstrate the added value of component video in its many scan rates and aspect ratios. Initially, price has to come second, and probably a distant second for some years to come. We recognize that CEA and the ATSC have a significant challenge in encouraging the general population to buy into this new system. They may be tempted to put price ahead of performance. We understand their needs, but would suggest that it is at least equally important to find a way to encourage high quality products. It is important to come up with a certification system that not only encourages consumers to buy into the system, but upgrade, supporting the expanded capable that broadcasters are being asked to deliver.
DTV isn’t easy to define. Some in the industry believe it’s a replacement for our analog transmission system. That’s only the beginning. Probably the most compelling reason for dropping analog in favor of DTV has nothing to do with its origins. DTV offers far more efficient use of each TV channel than is possible in the analog domain. The only near term requirements of broadcasters should be to go on the air with digital and source any video format in component. Why video in component? Conservation of bandwidth. It is a contributing factor, as is progressive versus interlaced. It will take some time for broadcasters to get there. Equipment availability for progressive video production is still a problem.
Many broadcasters currently transmitting DTV choose to waist that space by upconverting conventional resolution pictures to higher rates, requiring all of the digital space allocated to the new system. What should be happening is that standard resolution signals be broadcast at 480p, using the rest of the space for something else. On the program production side DTV has seriously expanded the canvas with which the artist can convey a message. How much do we want to short change that expanded message in a receiver by limiting its capability?
The ATSC has defined 18 to 36 scan rates and two aspect ratios, depending on who’s counting, so that the broadcaster can pick the right rate and shape for the program being transmitted. The FCC has wisely chosen not to limit the system to so few scan rates or aspect ratios. After all, in many minds, the ATSC left out at least 540p and 600p, let alone a common version of 480i. As the quality of MPEG compression improves we might be adding 768p to the list. If we add the Europeans to the mix we also need to consider 576p.
For those asking about 1080p programming, we certainly don’t expect it to be a viable format for distribution in the consumer world. The cost of a display device that can take advantage of this format is and will be out of the range of most high end consumers for a long time to come. Producing in 1080p is another story. It’s called protecting your assets for the future.
The original ATSC system called for two aspect ratios, 1.33:1 and 1.78:1. Maybe 0.77:1 is correct for program information that is in the shape of a piece of paper. You’ll find many people in the motion picture industry that will argue for all sorts of other aspect ratios. Among them, 1.37, 1.66, 1.85, 2.0, 2.2, 2.35, 2.39, and 2.4. Have I left any out? Probably.
Considering all of the possible option in digital transmission, is it probable to come up with one specification for DTV certification that would cover it all? What happens if we limit it to the more confined scope of the ATSC rates and aspect ratios? Either option presents a challenge to set manufacturers.
What’s really needed is a division of certification specifications. In figuring out what those categories should be, we might do well to go back to the 1980’s proposed analog transition from standard definition TV to HDTV. You may remember that was back when our high definition systems was going to be analog and maybe even compatible with the current NTSC system. (Digital signal processing has made that dream of backward compatibility possible.)
Following the analog lead, the new certification system would start at a Standard Definition (SDTV) level. It would include the 480i system we now use for the majority of our analog transmission. The interlaced PAL system, 576i, might also be included. The next step up would be Improved Definition (IDTV). We’d put 480p into that category. The next step up would be Extended Definition (EDTV). It would include a range from 540p, to say 720p, encompassing a progressive version of PAL and 1080i. Near the top quality category would be High Definition (HDTV). It would include 1080i, 720p, 768p and maybe go higher. Since we see 1080p being out of reach for all but a handful of consumers in the next few years, that rate would belong to a Super HD specification.
Is this all we need to know when defining the quality of a display device? Could we build a certification system for TV sets based solely on their scan capability? No. Other factors, such as horizontal resolution, proper decoding of the component video signals, gray scale tracking, the aspect ratio of the display itself and correct colors of red, green, and blue are equally important. Here’s where things get a little more difficult. Let’s take displayed picture resolution in particular. Horizontal resolution is probably the same for 480i and 480p, just as it could be equal for 1080i or 1080p. Does the p versus i look enough better to place the p in a higher category? Absolutely! The vertical resolution in the p version is enough better than the i signal to bump the p up to the higher category.
How important is horizontal resolution in any of these categories? There are a number of considerations to be taken into account when answering that question. We feel that horizontal resolution isn’t as important as it is credited in most circles. First of all, higher resolution comes at a cost. Light output capability is inversely proportional to horizontal resolution. As horizontal resolution goes up, light output goes down. It will go down to a point where if we were to attempt to clearly reproduce 1920 lines across the width of a 32 inch wide screen, the set won’t produce enough light to be considered for any certification.
The same light output versus resolution consideration isn’t necessarily true for CRT based projectors. Often times higher line rate capabilities in the source signal will produce better performance in the CRT. Yet there are significant limitations here as well. As the CRT size gets larger and the focus capability gets better, the performance increase is in picture detail, not more light output as manufactures would have you believe. You’ll have to keep the screen size down, even on 9 inch CRT type projectors, in order to obtain the performance capability of the tube. While we’re on the subject of picture performance from a projector, the screen material plays an equally important roll in picture quality. That must also be part of any real DTV certification specification. Any of the fresnel-lenticular screens we’ve encountered, including the super expensive screens we’ve seen at resent CEDIA, NAB and CES shows, would not fit into the ED or HD categories.
You might think that we’d have to consider a separate category for CRT based projectors in either a front or rear screen configuration. In reality consumer priced direct view sets, say below $15,000., wouldn’t qualify for the HD category, so projectors have that category to themselves.
While still on the subject of picture resolution, there are additional factors to consider in a direct view display device. The distance of the viewer from the picture is important. Anything more than one or two picture heights away from a 30 inch wide set and you’ll be hard pressed to see half of the 1920 lines no matter what the capability of the set. The conclusion here is that while horizontal resolution capability can’t be ignored, it’s not critical that it even approach the capability of the high definition source signals for most of the categories.
As for the matrix decoding, gray scale tracking and the primary colors of the display, those factors become more significant as you get beyond SDTV. They belong near the top of the list for consideration in the advanced categories. Since the matrix is important to color quality we should also cover the other two important parameters of color fidelity, the primary colors of red, green and blue, and gray scale tracking.
There is a set of primary colors defined for the higher resolution capability of the DTV system. It is slightly different from that defined for standard definition. That factor is almost unimportant as the standard definition color set hasn’t been widely available in the consumer market. Certainly at an ED or HD level, the color of red, green, and blue should closely conform the to SMPTE definitions for the 1080i and 720p systems. Fortunately they are the same. We’re including EBU colors for an additional reference because they are so close to the HD color specifications.
SMPTE C Colors
|SMPTE C Red||0.635||0.340|
|SMPTE C Green||0.305||0.595|
|SMPTE C Blue||0.155||0.070|
SMPTE 274M and 296M HD Color Points
EBU Agreed Upon in 1969
As for gray scale tracking, it’s as critical as the correct colors of red, green and blue, and must be easily calibrated. What does that mean? Access to the controls should be easy for the person trained to properly set gray scale. The resolution of the controls should be good enough to make accurate calibration easy. Once the set is calibrated at the high and low points of the gray scale, it should also track gray within a certain specification. The number commonly used in the broadcast world is 6 CIELUV. That’s three times the minimum perceptible difference in color. We would assign specifications to DTV categories according to their position on the quality scale.
What about the shape of the set? SDTV and IDTV could both be 1.33:1, with the IDTV set having the capability of displaying a 1.78:1 image somewhere in the 1.33 area of the display. Display devices in the EDTV or HDTV categories should be widescreen, a 1.78:1 aspect ratio if we stick to the ATSC criteria.
There are several more important points that should be spelled out prior to certification. The display device can easily be limited by its input(s) and signal processing. Where an external tuner is required we might only be capable of declaring a display device as DTV ready. If a DTV tuner is part of the package its capabilities will also have to be considered. Then there’s audio. Part of higher end certification for display devices with a built in DTV tuner might include a serial digital audio output for external processing of the audio. In any event a high end display device should not be part of the acoustic reproduction capability.
Looking at the video input, it is our experience that a Y Pb Pr connection is critical to obtaining good picture quality, plus having an advantage in setup. A black and white signal is needed to help assure proper calibration of black and white levels, gray scale and convergence. That’s easy to obtain if you can disconnect Pb and Pr of the component input. If the input to the set is serial digital, an internal capability of shutting off the Pb and Pr must be provided. If the display device only provides for an RGB input, as is the case with most high end video projectors, a Y Pb Pr to RGB converter must be part of the specification for high end DTV certification. The bandwidth and matrix of such a device must conform to SMPTE 274M and 296M.
We’ve already alluded to the fact that the DTV system is dynamic. Specification set today will most likely have to be changed in the near future. As an example, the standards for data transmission have yet to be set so we can’t specify how a current generation set is to handle that part of DTV. We might even have to date the specifications, calling them SDTV 2000 or HDTV 2002. There is another reason for wanting these specifications dated. It is our firm belief that in the future all display devices will run at their own rate, and at just one rate, totally independent of incoming signals. We feel that the cost of high quality translation from any incoming format to the ideal rate for the individual display will be far less than building a multisync or multiscan display device. This approach is required for fixed array displays that are currently on the market and will be beneficial if used on display devices capable of several rates.
The DTV tuner specifications are obviously important to the resulting picture quality displayed on any of these sets. A separate certification process will have to be set up for that category. We are not addressing that subject in this proposal.
Now you have the background of this discussion. Let’s see what happens if we apply what we’ve talked about. We’ll let you know that what you are about to read is the fifth draft of this document, based on input we’ve received from past editions. We fully recognize that our proposals of the specification for EDTV and HDTV aren’t necessarily going to happen in current stand alone equipment. We do feel that combinations of equipment could be assembled that would meet these specifications. Comments are welcome.
SDTV, Standard Definition Television
Since these are “DTV Certified” specifications for SDTV the criteria is going to be more stringent than might be found from a regular TV set, yet not much beyond the current capability of good sets.
Scan rate capability need only cover our current analog interlaced TV system of 525 lines at 59.94 Hz otherwise know as 480i in the DTV system. All color decoder and sync specifications apply for being able to handle VHS in standard play as well as fast forward and rewind. Should the PAL rate of 625/50 be considered? That might depend on the market, but is not necessary in the U.S. market.
A display aspect ratio of 1.33:1 would be acceptable, but not limited to that shape. An aspect ratio of 1.78:1 would also qualify. What about other aspect ratios? How much unnecessary confusion do we want to add to DTV? None is the correct answer. We would only consider 1.33 or 1.78.
This set would not be required to have multiple aspect ratio capability in the first generation, relying on the DTV set top receiver to do the aspect ratio conversion.
In a few years the SDTV DTV certification category would be upgraded, requiring a 1.78 aspect ratio display capability. Scan rate capability would increase to a minimum of 480p; with a built in i to p converter for 480i sources. The scan rate converter should have 2/3 recognition capability for film sources but would not be required on the first generation. A 1.33 screen shape will continue to be acceptable as long as the raster could be sized for a 1.78 shape.
Analog Y Pb Pr input(s). A flat Y frequency response out to 7.0 MHz ± 0.5 dB. A Pb and Pr response out to 3.0 MHz ± 0.5 dB.
Accurate translation (decoding) from Y Pb Pr to RGB using the NTSC matrix
The RGB path response inside the set should be flat out to 10 MHz ± 0.5 dB. Any “Sharpness” function would have to have a bypass mode. This means that in the by pass mode there will be no sign of the phase errors that are often introduced by the presence of sharpness circuits.
For compatibility with conventional analog signals the set must contain a composite and S-Video input with an accurate color decode capability. An adaptive comb filter would be required in this application.
Any SVM circuits incorporated in the circuit design must be easily shut off for the DTV display mode.
Easily calibrated gray scale, capability of tracking D65 to at least 12 CIELUV. The resolution of the calibration controls should be good enough that the calibration points can be set within 1 CIELUV of the target color of gray.
Display light output capability of 25 ft-Lamberts without blooming.
The choice of colors for red green and blue will most likely be left to the manufacturer for the first generation of sets in this category. We would suggest something close to SMPTE C colors in the first generation and insist upon it or the HD color set for the second generation of the SDTV specification.
Displayed horizontal resolution of 530 lines across the full width of the screen. This translates to about 400 horizontal lines per picture height for a 1.33 picture; the analog TV specification for horizontal resolution. We are using a full picture width in our specification because it is independent of aspect ratio. Inexpensive TV sets can now do from 380 to about 420 lines/picture height, yet we find only a few expensive sets reaching as high as 450 lines. (This despite claims of 700, 800, or even 900 lines of resolution on the part of some manufacturers.)
Separate setup for brightness, contrast, and gray scale for the composite or S inputs versus the component inputs.
IDTV, Improved Definition Television
Display scan rate capability starts at 480p. While the set will be able to accept lower rates, they must be upconverted for display. This set may or may not include a PAL progressive capability. It would not be required to go any higher than either of these two rates.
A minimum displayed horizontal resolution of 560 lines across the full width of the screen.
An aspect ratio of 1.33:1 would be acceptable in the first generation, but not limited to that shape. An aspect ratio of 1.78:1 would be encouraged in the first generation of this specification and mandatory in the second generation of the specification. An additional capability of displaying a 1.78 image inside the 1.33 picture area, if the set is a 1.33 aspect ratio would be required for the first generation. If the set is a 1.78 aspect ratio we would require the capability of a 1.33 image in the center of the 1.78 image area. In addition there would have to be a capability of displaying a letterboxed 1.78 image out to the full width of the screen. Individual brightness and contrast memories would be required for each aspect ratio displayed.
Y Pb Pr input. RGB or VGA is optional on the first generation. The Y Pb Pr decode matrix would be NTSC. A flat Y frequency response out to 7.0 MHz ± 0.5 dB and a Pb and Pr response out to 3.0 MHz ± 0.5 dB for the 480i input. If the component input could also accept a progressive signal, the frequency response specification would have to be doubled. The RGB or VGA input response would be flat in each channel to 20 MHz ± 0.5 dB all the way to the imaging device.
For compatibility with conventional analog signals the set would contain a composite and S-Video input. Accurate color decoding of composite and S video would be required. A high quality adaptive comb filter would be mandatory for going between composite and s video. The set would contain an internal processor to convert the 480i signals to 480p. This same processor could be used to convert any DTV 480i source to 480p. That means the processor would have a component as well as an S and composite input capability. The component path would conform to the bandwidth requirements stated in the SDTV specifications. We would encourage the manufacturer to include 2/3 pull-down recognition in this video processor in the first generation with it being required in the second generation.
Matrix decode capability for 480 i and 480p only. (They are the same thing.) Conversion from higher rates, including the change in matrix, is to take place outside this set. Individual memory would be required for component sources originating in 480i and 480p. It is likely to run into 480i sources with setup on black and 480p sources without setup.
Capability of turning any SVM circuits off in any display of DTV signals.
Easily calibrated gray scale, capability of tracking D65 to no greater than a 10 CIELUV variation from 0.5 ft-Lamberts to full brightness. Resolution of the controls at the calibration points should be less than 1 CIELUV. A minimum of a second choice of color temperature must be made available. It needs to have a range that could easily reach 5400° Kelvin. The color temperature memory would be selectable from the front panel or remote control. It needs to be assigned to individual inputs memories so that when the user selects a particular input, the correct color temperature will also be selected. We will encourage at least four memories for color temperature, any of which can be assigned to a particular input in the first generation, requiring it in the second generation.
Light output capability of 25 ft-Lamberts without blooming.
Separate setup for brightness, contrast, geometry, blanking and gray scale controls for each displayed aspect ratio. There would be a minimum of two choices for gray scale, each of which could be individually calibrated and assigned to particular memories. What would be included in geometry? Certainly height and width plus any other controls needed to insure reasonable geometry at each aspect ratio. In a direct view set, it is our experience that separate vertical linearity memory would be necessary for the two display aspect ratios.
EDTV, Extended Definition Television
In the initial stages of specifying this set we would call for a multi-scan capability in CRT based display devices. It would start at 480p and run up to 768p. That range would include 1080i which is just above 480p. Careful attention would have to be paid to the ringing in the picture caused by the short retrace time of the 1080i signal.
It is our feeling that down the road ID and ED set will only scan at a single progressive rate. Second or third generation of specifications for these categories would include an internal scaler that would convert incoming signals to the specific rate required by the display.
Fixed array display devices entering this qualification would now have to have processors that would convert the incoming signal to the display configuration of the device. Processor quality specifications would have to be included.
Input connections would include a composite 75 ohm BNC for NTSC and PAL, an S-Video connector, a component input on three 75 ohm BNC’s, and RGB plus H and V sync on five 75ohm BNC’s, plus a VGA type connection. We would encourage at least two each component and RGB connections. The two RGB input connections could be divided into one for the BNC connectors and a second at the VGA connection. The DVI connection could be part of a first generation specification and probably mandatory on the second generation.
The set would contain an internal signal processor for 480i source signals. Conversion could be to any rate between 480p and 768p. The converter would be required to recognize 2/3 pull-down in film original material. The NTSC and S video decoder would have to be color accurate. We would require at least a high quality 2D adaptive comb filter and would encourage a 3D adaptive comb filter. The output of the second generation of this converter would have to run at the ideal rate of the set.
The set would have to contain a matrix decode capability for both 480 i and p signals and the higher rates as specified by SMPTE 274M and 296M. We would encourage automatic detection of the input rate for the proper selection of the decode matrix with a manual override capability. The equation for Y in these SMPTE documents is Y’ = 0.2126 R’ + 0.7152G’ + 0.0722B’
An ability to shut Pb and Pr off for component video calibration would be part of the setup menu system. An ability to display blue only for NTSC decoder calibration would also be included.
Luminance bandwidth of the entire component and RGB path would have to be flat out to at least 40 MHz, ± 0.5 dB. Pb and Pr response would be flat to 20 MHz ± 0.5 dB. Any sharpness function that might be included would have to have a bypass mode such that it would not introduce any frequency or phase response errors in the 0.1 to 35.0 MHz band.
If a frequency detection method is used to differentiate between the 480 i and p rates from the higher rates, the 480i signal path into the processor could be band limited to 7.0 MHz for luminance and 3.5 MHz for the two color channels. The 480p signal would not be band limited, as is optional for the 480i input, but sent to the proper decode matrix. The circuits might be kept simple if the line processor, converting 480i to a progressive signal, were done in the component domain. The output could them be fed to the 480p matrix decoder.
The red, green and blue colors of the display would closely conform to SMPTE 296M specifications for 720p. (That same specification also applies to 1080 i and 1080p.)
The shape of the picture would be 1.78:1
The set would contain full multiple aspect ratio capability. We would insist on four but encourage a minimum of six options of memory for aspect ratios. We would require good geometry setup for a 1.33 picture in the center of the 1.78 display, an overscanned 1.33 image, a linear 1.66 aspect ratio; letterboxed, a linear 1.78 image; letterboxed and a 1.78 image edge to edge of the raster, sometimes know as the anamorphic format. It’s this 1.78 edge to edge position that would most likely be used for the 1080i and 720p rates. Memories for each would have to contain individual information for brightness, contrast, color temperature, and picture shape and geometry.
Separate memories would have to be provided for each input rate. As much as there would be multiple aspect ratio capability, it would exist individually for each rate. It would also exist individually for each input connection on the back of the set.
Monitors that are 32 inches wide or smaller would provide 25 ft-Lamberts of light output prior to any noticeable blooming in the picture. Flat field uniformity would be within 12 CIELUV for color and no more than 45% fall-off from center to edge in luminance. These measurements would be checked over a 60° angle from the center of the picture, both horizontally and vertically.
Light output could drop to as low as 10 ft-Lamberts as the screen width approached 72 inches. The 12 CIELUV and 45% fall-off specification would still apply to the larger screens. Measurements would be made over a 60° angle from the center of the picture.
Gray scale tracking would have to be within 10 CIELUV, with a resolution of less than 1 CIELUV at the calibration points.
Horizontal resolution across the width of the screen would have to be at least 800 lines for a screen size of 27 inch wide and approach at least 1200 lines in a 72 inch wide image.
We don’t yet know the numbers for quality of geometry and convergence, but need to convey that they would be tight. We expect them to be in the order of 0.3% of the total dimension.
HDTV, High Definition Television
Input and processing requirements of EDTV. We would extend the higher scan rate input bandwidth to being flat out to 60 MHz, encouraging and internal RGB bandwidth to be flat out to at least 80 MHz.
Scan rate capability would include 1080p although we would not expect the display device to fully resolve the horizontal bandwidth of 1920 lines across the full width of the screen.
Horizontal resolution would be specified as a minimum of 1600 lines. If a “Sharpness” function is included as a “feature” it would have to have a bypass mode where no frequency or phase errors were introduced in the 0.1 to 60 MHz portion of the video path.
Light output capability would be placed at 10 ft-Lamberts for a 72 inch wide screen 1.3 gain screen with no visible blooming in the image.
Display quality for color uniformity would parallel the EDTV set. Luminance uniformity would be within 30%.
If you thought the specifications for geometry and convergence were tight for the EDTV category just wait until you see what we come up with for this category. We’re currently looking at geometry errors no larger than 0.1% of the picture dimension.
We mentioned this in the article. We don’t now have many display devices that can truly show off the capability of a 1080p signal, let alone the 2K by 2K or 4K by 3K images that are now being created in the graphics industry. It is our feeling that a 12 inch electromagnetically focused CRT projector would be required for this task. While we’ve seen such projectors, none of which were properly designed for such high resolution video display. A requirement for this type of display is here now because the signals are being generated. It may happen in another technology before it happens in the CRT world.
We believe that anyone thinking that good HDTV can be done at an average consumer price, any time soon, is pushing us head long into the HD version of “Never Twice the Same Color.”
Politics of HDTV
We brought this up so we may as well address it, although briefly. It is our opinion that broadcasters have a right to expect that consumers should have a choice of high quality display. This might even be necessary prior to broadcasters being expected to take on HDTV for all prime time programming. For the time being good display devices will be slow to come. They will cost a lot of money and only a small percentage of the population will be able to afford them. That does not mean that any delay should be allowed in the implementation of DTV transmission, just a possible relaxation of prime time HD requirements. We feel the 2006 cut-over date should be absolute, independent of how many consumers own set top boxes or good DTV sets.
If TV set manufacturers or any other advertisers wish to continue to fund network presentations of HD material, that is their choice. Consumers and broadcasters will welcome it. It should help to sell the expensive, profitable HD sets.
We would suggest that real HD get started as a DVD format, where quality standards could be set higher than anything we would expect from broadcasters or DSS sources. That is a world where everyone involved can make money. It should be used to create the base needed of HD displays to attract advertisers to the networks.
We would certainly like to see local TV stations shift to at least 480p, 1.78 aspect ratio, digital component video production as soon as possible, but don’t feel they should be obligated by Congress or the FCC to go any further until TV set manufacturers step up to the plate with real DTV sets.
Joe: Kane Productions is located at
Joe Kane Productions
12526 Otsego Street
Valley Village, CA 91607